Methodology for Content Analysis in Text Mining Projects and Fifth Generation of CAQDAS

Abstract

Projects which we work with—large volumes of text data that are acquired from various sources and stored in a variety of formats—rise many dilemmas of a methodological nature, often require unstandardized decisions and solutions. In particular, compiling data of various quality, unstructured types, and of quan and qual nature requires dynamic strategies, ideas, and ways of analysis. The article describes an example of this approach. It shows its value in classification of written statements. In such context, the author of the article faces the following objectives: (a) can we use CAQDASso that semiautomatic or automatic work would replace some manual work regarding classification of the expressions; (b) how to construct a classification methodology for data of various quality; (c) when the automatic classification is useful and when there is no chance of success? The article will be marked with moments in which the analyst reaches for knowledge typical for qualitative data analysis, and when the knowledge of this area is no longer sufficient to classify content (natural language processing, machine learning). An example of a project being the background of this article forced the use of several tools and languages to support work with the data. Work on the transformation, classification, and visualization of results required applications such as: MySQL, R, QDA Miner, WordStat, Qlik Sense. Role and limits of the computer-assisted qualitative data analysis software tools have also been noted.